GB2249238A - Image encoding apparatus - Google Patents

Abstract

Image data, 7, for encoding is processed by a plurality of block truncation coding means 3, 4, 5, each of which encodes a proportion of the elements of each block (Fig 9) according to the number of gradation levels. A judgment unit 1 determines the tone gradient of each block of incoming data and thus determines whether 1-bit encoding of a large number of elements (e.g. all elements, fig 9A and encoder 3) or m bit encoding of a smaller number of elements (half or a quarter, Figs 9B, 9C, and encoders 5, 6) should be selected. As described the number of data bits transmitted per block is constant. In an alternative embodiment (figure 10) the block judgment or classification unit 1 calculates a distortion measure from the block encoder outputs by a square error average or absolute error average distance. <IMAGE>

Description

2249233 4 Maz mmoniza AREARL= BAMM= OF THE -INVENTION "rhe, fresixt

ttiverbi v^ Telates to CL M f -t6re 07 continuoizytj tone image encoding apparatuc.

in a facsimile terminal for inatancer original Imag signals are encoded In order to improve the transmission efticiency thereof. Particularlyr the amount of A=ge data having a hall tone or continuous tone Is quite larger and therefore highly efficient coding is required.

An example of such efficient coding has been proposed as a so-called BM (Block Truncation Coding) scheme in Published Unexamined Japanese Patent Application No. Sho-57-174984, In which an image signal is divided into block units each having a certain size, and the shape of every block is to be truncated.

The BTC coding scheme will be described with reference to Pigs. 1 to 5. In the ETC coding schemer one picture which is as shown in Fig. 1, is divided into a plurality of block units each being composed of Li x Lj picture elements as shown in Fig. 2. Assuming that L - Li= L, and that each tone of the picture elements in a block is represented by aLj, then the average tone PO of the whole of the block ia Po = E aij/L2. As shown In Fig. 3r if the averago tone and the number of picture elements having lower tone than the average tone PO in the block are P, and NI, rcar."ctivc>ly, then the average tone P., and the number NI are expressed by., L L P, - E E a:LjIN, l NI = Ect j j i j a,,j.:SP.

if the average tone and the number of picture elements having higher dencity than the ave=age tone Po in the block are F2 and N., respectively, , then the average density P2 and the number N2 are expressed by:

L L P2 E aij/N2, N2 'c ETL3 j ajAP, These expressions are established under the condition of 4,Lj = 0 it a:Lj Ii P,,, and Lj = 1 if aij > Po.

Now, in case of inserting integers 1 and n smaller than L2 and the number of gradation levels of tone, it is discriminated as tone distribution density in the block being uniform that 1PI-P21 < m, N1 < n 0= N2 < nt so that each bij is made zero. AS shown in Pig. 4, the whole picture elements in the block are represented only by the average tone PO. It Is discriminated as tone distribution density in the block being not uniform that 1PI-P21 a m# NI k n and N2 Z n, so that the block is represented by the both average tone P, and P, as shown in Pig. 5. The picture elements where oblj Is made zero &=& rep=oontd by P, whmraaa tho picture elements where chis Is made 1 are represented by P?. In this caset ij represents the Information relating to the shape of the blocki raferrod to as resolution information, while P0j, P, and/or F2 represent the information which indicate the tone gradation# reforred to as gradation information. The resolution information is encoded every several pair of lines en bloc by an ordinary binary coding system, and the successive length of blocks equal to each other in the information values is also encoded by a well known run length coding scheme to be transmitted. As to parameters;D and n in this coding system, ja serves as a discrimination of threshold to eliminate an isolated noise in an image, and p_ serves as a discrimination threshold to eliminate a small fluctuation of tone in a block, so that the image is to be smoothed so much as both m and n have larger values.

In this scheme, however, since picture elements in a block can be expressed only by two specified tone levels at maximum, there has been an issue that if block size L is made larger to obtain higher corapressiont the image degradation becomes large, and particularly the distortion is not negligible in the area having a smooth tone variation. In addition, there is much redundancy, since the resolution information in allocated uniformly to all blocks. Although in order to address this issue, it has been tried to reduce th rodundancy by means of the binary coding of the resolution information every several lines en bloc, it has not been satisfactory.

in addition, there has been a problem that it is difficult to control the coding rate by xeleotlng the parameters M and D_, and it is also enable to edit an image as it is in an encoded form.

SUMMY OF MIE INVB=ICW In order to eliminate the foregoing problems accompanying the conventional one, the present invention has been accomplished, and an object of the invention is to provide a coding scheme in which performed in such expansion that blocks can be expressed by using a plurality of different numbers of gradations in accordance with the contents of an image, and only required resolution information corresponding to the used number of tone level is transmitted so as to preserve the image quality even in the case where the block size L is enlarged to obtain more compression ratio, and the coding efficiency is made larger because no redundant resolution information is transmitted.

The above, and other objects of the present invention are accomplished by the provision of an image encoding apparatus comprising: means for sampling an image to divide the imaqe into a plurality of blocks each being composed of the same number of picture elements; a classifying means for classifying the blocks in view of the degree of tone gradient of each block; and a plurality of block truncation coding means allocated so that each of the blocks has a predetermined code length of gradation and resolution therein, the plurality of block truncation coding meant being suitably switched in accordance with the result of classification by the classifying raeana to thereby enable block truncation coding.

According to the present invention, blocks of an image can be expressed by use of a plurality of tone levels, for example, from one or two tone levels. to m, M being two or more,, and resolution information which is required in view of the used number of tone levels is only transmitted to maintain the excellent image quality, even in the case where the size of block to be encoded is made larger to increase the coding efficiency. In addition, as known as Weber's law, a man has such a human visual characteristics that it is unlikely to have low sensitivity in tone level in an image area having a large tone variation, while it is likely to have low sensitivity in resolution in an image area having a smooth tons variation. According to the present inventiont therefore, the number of gradations is increased in an image area having a smooth tone gradient and then the resolution of image in made lower. On the contrary, the number of tone levels is decreased in an image area having a large tone variation and then the resolution is made higher. It is therefore likely to visually neglect the degradation of image quality caused by encoding, and no redundant tone level data and resolution information are zubjected to encoding so that the coding efficiency is improved.

MIEF DESCRIPTION OP TW DaLW_INGS in the accorapanying drawinget rig. 1 is an explanatory diagram lllust=ating a picture divided into a plurality of blocks; Fig. 2 is an explanatory diagram Illustrating the structure of each of blocks; Fig. 3 is an explanatory diagram illustrating the average tone used as a reference in coding; rig. 4 is an explanatory diagram illustrating a state of coding in the case where the tone variation of an image is smooth; Fig. 5 is an explanatory diagram illustrating a state of coding in the case where the tone variation of an image is large; Fig. 6 is a block diagram showing an embodiment of an image encoding apparatus according to the present invention; Fig. 7 is a block diagram showing an example of a plurality of encoders employed in the image coding apparatus of Fig. 6; Pigs. 8(a) to 8(c) are explanatory diagrams illustrating states of quantization in the encoders shown in Fig. 6; Figa. 9(a) to 9(r-) are explanatory diagrams Illustrating the difference in resolution information among the encoders; and rig. 10 is a block diagram chowing another embodiment of the image encoding apparatus according to the precent. invention.

QMS-CRIPTION OF THE FREMMD MMODMWS An embodiment of the present invention will be described hereinafter with reference to the accompanying drawings.

Fig. 6 is a block diagram showing an embodiment of an image encoding apparatus according to the present invention.

First, the operation of the image encoding apparatus according to the present invention will be described briefly under an assumption that inputting operation of image data to be encoded Is carried out every such a block as shown in Pigs. 1 and 2, for example, composed of L x L picture elements.

A block classification unit 1 classifies the degree of tone gradient In a block according to the image data 7 inputted to supply a classification result 14 to a selector 6. In the embodiment of Fig. 6, an encoder 3 performs single tone block truncation encoding, an encoder 4 performs dualtone coding, and an encoder 5 performs m number of tone level coding. in case of the degree of tone gradient being large In the block, the outputs which are produced by encoders for small nunbe= of g=adation, such as encoders at the side of encoders 3, are mainly employed by the selector 6. on the other hand, In case of the degree thereof being s=oth, the outputs which are produced by encoders for large number of gradations, such as encoders at the side of encoders 5, are mainly employed by the selector 6. As will be described later, as to resolution information, coding is performed such that many cases are allocated to the encoder 3 which performs single tone block truncation encoding while a few cases are allocated to the encoder 5 which performs n-tone-level block truncation coding. In case of a large tone gradient, therefore, coding is performed under the conditions of a small tone information and much resolution informationy and in case of a smooth tone gradient, coding is performed under the conditions of much tone information and a little resolution information. The selector 6 selects and outputs the data 10-1 to 10-3 which are encoded by any one of the encoders 3 to 5 in accordance the block classification result 14.

Ne xt the operation of the above-mentioned image encoding apparatus will be described in detail.

In this wabodiment, an image Is divided into 8 x 8 picture elements block and the Image data 7 shown in Fig. 6 in inputted by every block of a x 8 picture elements. In addition, three kinds of block truncation encoding, that let dual tones encoding, tour-tones encoding and sixteen-tones encadingi are porformed adaptively. The number of Picture elements in one block may be changed and the shape of blocks may be modified.

As hag been described abovQj the block classification unit I operates to classify whether the tons gradlent of picture elements in a block is smooth or complex, the tone variation of the same is large or not, or the tone distribution is intermediate or not. Particularly, the block classification unit 1 measures the tone distribution of picture elements in a block. According to the classification result, the unit 1 classifies the tone variation to-be sharp when the dispersion is larger classifies the tone gradient to be smooth when the dispersion is small, and classifies the tone distribution to be intexmediate when the dispersion is intermediate. The above same classification may be accomplished by differentiating a picture element tone value with respect to a spatial domain, so that the numbers of the maximum and minimum values may be obtained as criterion.

Since the respective encoders 3 to 5 have the same basic circuit construction, they will be described in common with reference to Fig. 7.

Upon reception of the image data 91 a quantization level controller 15 determines quantization levels Ql, 02r -- on the basis of the received image data 9. The number and values of these quantization levels are different among the respective encoders: for example, the encoder 4 for perfo=ing dual tone block truncation encoding, as shown in Fig. 8(a), has two quantization levels Ql and Q2. in the caae of four or sixteen tons lovalar as shown in Fig. 8(b) or B(c), there are four quantization levels Ql to Q4 (Q2 and Q3 not shown), or sixteen quantization levels Q1 to 016 (Q2 to Q15 not shown), respectively.

More specificallyt as shown in rigs. 8(a) to 8(c), the maximurn quantization level Q1 and the minimum quantization level Q2, Q4 or Q16 are determinedr and the difference between the maximum and minimum levels is equidistantly divided so as to perform linear quantization. The maximum quantization level Q1 is an average value of the tone values of n image elements sequentially taken in the descending order from the maximum value "max" of the densities of picture elements in the block. In the same manner, the minimum quantization level is an average value of the tone of _n picture elements sequentially taken in the ascending order from the minimum value "min" of the densities. Assuming that the number of picture elements in the block is L2 and the number of gradations is L. then the parameter IL is truncated as a value of order expressed by n L2/1 - The maximum and minimum quantization levels may be used az they are. in addition, not limited to the case of the linear quantization between the maximum and minimum quantization levels at equal steps, tor example, nonlinear quantization such as quantization of MAX may be performed.

That ir., on the assumption of a stochastic tone function, a quantizer optimum thereto may be designed.

Next, a gradation data encoder 16 encodes a quantization level 11 supplied from the above-mentioned quantization level calculator 15. Specifically, as shown in Figs. 8(a) to 8(c), the gradation data encoder 16 outputs the average value La of the maximum and minimum quantization levels Q1 and 02 or Q4 or 016 and the difference Ld therebetween. In this embodiment, under an assumption that the input inage data 9 has 256 gradations per a picture element and is expressed with 8 bits per picture element by binary expression without sign, the La and Ld shown in Figs. 8(a) to 8(c) are expressed by the number of gradations equal or nearly equal thereto. That is, the La and Ld are expressed with 16 bits and the like. The manner of encoding the qtantization level 11 is not limited to the above method, and the maximum and minimum quantization levels may be allocated as they are. In addition, in the case where non-linear quantization is performed in the quantization level calculator 15, the way of change of a quantization stop (the difference between adjacent quantization levels) must be also encoded.

A quantizer 17 quantizes the above-mentioned input imago data 9 in accordance with the quantization level 11, and outputs the result as resolution information 12. In this case, the threshold value of the quantization is an average value of corresponding quantization levels. For example, the threshold value between the quantization level$ Q1 and Q2 is expressed by (Ql+Q2)/2. 3:n thix embodimentr the quantized resolution information 12t as quantized into twOr fOur Or sixteen gra dations, is one-bit, two-bits or four-bits datar respectively.

Next, a resolution information 5010ctOz IS eliminates visually redundant data from the resolution information 12 correspondingly -to the number of gradations quantized by the quantizer 17, and outputs required resolution information 13. Tn this embodiment, as shown in Figs. 9(a) to 9(c), all the resolution information of 64 picture elements (- 8 x 8 picture elements) is selected in case of dual tone encoding. In case of four-tone levels encoding, resolution information of 32 picture elements (= 8 x 8 picture elements/2) is selected in a quincux form. Further, in case of sixteen-tons levels encoding, and resolution information of 16 picture elements (= (8 picture elements)/2 x (8 picture elements)/2) is selected in. an orthogonal form.

In Fig. 9, the hatched portion shows the selected picture elements. How to select resolution information is not limited to the above example, and for example, the resolution information of 32 picture elements may be selected in a quincux form in the case of dual tones encoding. Here it is important that if much gradation data is allocated to a certain block, resolution information is allocated a little, and on the contrary, if a small gradation data is allocated, resolution information is allocated much.

Next, a resolution info=&tLon ancoder 19 OPOratffis to. encode the abovementioned selected roool,tion jnELITation 13. In this embodiment? without using encoding for restraining tone and/or resolution redundancyr the encoded image data 10 is outputted by expressing the zelected resolution information 13 in binary number without sign an it is.

The selector 6 of Fig. 6 operates to select and output the encoded image data 10-1 to 10-3 supplied from the encodera 3 to 5 in accordance with the block classification result 14.

According to the above mannerr the imaqe data 7 supplied for every block is encoded in combination of tone level information with resolution informationt and the combined data is outputted as the encoded image data S.

Table 1 shows the relationship between the tone level information and the resolution information truncation coding upon respective gradations.

Tables-1 four- sixteen two-graduation graduation graduation block block block truncation truncation truncation encocting encoding encoding quantization level Qir Q2 Q, to 04 Q1 to Q16 (number of (1 bit) (2 bits) (4 bits) bits) tone level information La, Ld Tja, Ld lia p Ld (number of (8 bits x 2) (8 bits X 2) (8 bits X 2) bits) resolution 64 bits 32 bits 16 bits information coding 8 X 2 8 x 2 8 x 2 64 + 2 x 32 + 4 x 16 + 1 quantity x - 80 bits - bits - - 80 bits As shown in Table 11 in this embodimenty regardless of the result of block classification, constant rate of encoded data (80 bits) is always allocated and encoded for every block. If block selection data of two bits is added every block in order to identify which encoder has been selected, encoding efficiency or regulating ratio in this embodiment will be (8 x 8 picture elements x 8 bits/(80 bits 2 bits)) = 6.244.

In the tons level data encoder 16 and the resolution information encoder 19 in Fig. 7, the encoding for res-training both tone level and resolution redundancy, for example, Huffman coding, arithmetic codingr or tho like, is not used.

Nextf the procedure of decoding in this embodiment will be described briefly. First identification is made with the above- mentioned block selection data an to which one of the encoders 3 to 5 in rig. 6 is selected at the time of encoding, and the tons level information La and Ld and the resolution information shown in Fig. 8 are decoded. In this embodiment, since such encoding as Huffman coding is not used, it goes well if binary numbers arranged in the predetermined order and bit-length are read successively. Next by use of the tone level information La and Ld, the difference from (La+Ld)/2 to (La-Ld)/2 is divided at equal steps, so as to calculate the quantization levels Ql, Q21 --. The image data is reproduced on the basis of the obtained quantization levels and the resolution information. At this time, if certain resolution information is missed out selectively, for example, in the horizontal or vertical direction, interpolation is performed by use of tone values of neighborhood picture elements having been reproduced, so that the tons of the missed picture element is reproduced. Through the above procedure, an encoded image is decoded.

Fig. 10 is a block diagram showing another embodiment ol the present invention. It is different from Fig. 6 that the image data 10-1 to 10-3 which are encoded by encoders 3 to 5 at the same time are supplied to a local decoder 20, and a selector 6 is controlled by the output of the local decoder 20. The local decoder 20 decodes the encoded image data 10-1 to 10-3 to be compared with the image data 7 before encoding of the encoders 10-1 to 10-3 in a block classification unit 1. According to the rCault 14 of the comparison, the Golactor 6 operates to select and output the encoded imago data having the minimum distortion, measure. That lap the block classification unit 1 acts as a distortion measurement means. There Is no restriction as to the scale of the distortion? and a square error average distance, an absolute error average distance may be used.

In the second embodiment of the present invention shown in Fig. 10, it is not necessary to classify the tone condition of the block, that Is. whether an image inputted for every block has a tone gradation which is comparatively smooth as a whole of the block, or has a tone variation which is quite large as a whole of the block, or has an intennediate tone variation between the above two conditions as a whole of the block. in order to classify which one of the above-mentioned encoders can minimize the visual degradation of image quality upon coding the image data supplied for every block, the dispersion value of tone of picture elements in every block or the differential coefficient in a spatial domain is used in the embodiment shown in Pig. 6. However, It cannot be said that the Ideal classification is always performed. in the other embodiment shown in Fig. 10, on the contrary, the ideal classification is porio=d baa&d on at least a distortion measure such as a square exror average distance, accumulation of an absolute error average distance, or the like. This embodiMent is advantageous in that it ia possible to perform coding with less degradation in image quality.

As has been described, according to the present invention:

(1) Even in the case where the block size to be encoded is enlarged to improve the encoding efficiency, the degradation of image quality is small because encoders for two or more tone levels are used suitably.

(2) Since a plurality of encoders are used suitably, it is possible to accomplish the reproduction of an image having good quality regardless of the kinds of images. In other words, it is not easy to visually detect the degradation of image quality caused by encoding a variety of images such as a character/traces image in which resolution reproduction is visually focused on, an image such as a figure/scenery image in which tone level reproduction is visually focused on, and an image having an intermediate characteristic between the above two images. in addition, high coding efficiency can be attained without using coding such as Huffman coding for restraining information-thaoxetical redundancy.

(3) Since the coding efficiency is always made stable for every block without using coding such as Huffman coding for restraining redundancy, in the case where encoded inage data In stored in a nw=ry having a certain constant capacity, or in the case where the encoded image data Is stored In or rep=odu from a secondary storage device having a certain constant transfer rat# It is not necousary to control the coding etticiency (code quantity)i so that It is possible to reduce the capacity of a buffer memory or omit the buffer memory per se. In additionr since It Is possible to achieve Independent access, block by block# only a certain part of image data which is encoded image by Image, It Is possible to execute image edition such as cutting-out, transferring. moving, deletion, or the liker of an Image as it is in the state of encoded Image data, at a high speed.

i CLAM 1. An image encoding apparatur. comp=:Lsingt means úo= sampling an image to divide said image IntO a plurality of blocks each being composed of the am number of picture elements; a classification means for claasiúying said blocks In view of the degree of tone gradient of each block; and a plurality of block truncation Coding means allocated so that each of said blocks has a predetermined code amount of tone and resolution therein, said plurality of block truncation coding means being suitably selected in accordance with the result of classification by said classification means to thereby perform block truncation coding.

2. An image encoding apparatus according to claim 1, wherein said classification means comprises a local decoding means for decoding plural data obtained by simultaneously encoding said blocks by said plurality of block truncation coding means; and a distortion measurement means for obtaining trunction between said blocks and a plurality of local decoded blocks obtained by said local decoding meanso wh4gireby a classification result of said blocks Is obtained from the block having the minimum distortion among said plurality of local decoded blocks.

3. An Imge encoding apparatus according to claim 1, wherein said classification means comprises a local decoding means for decoding plural data obtained by euccowivolY encoding said blocks by said plurality of block truncation coding means; and a distortion measurement means for obtaining truncation between said blocks and a plurality of local decoded blocks obtained by said local decodIng means. whereby a classification result of said blocks is obtained from the block having the minimum distortion among said plurality of local decoded blocks.

Claims (1)

1. 4. An image encoding apparatus according to Claim lp wherein if data as to

   resolution information for picture elements is truncated in a predetermined direction selectivelyp the truncated data is Interpolated based on the decoded data of picture elements neighbocring therearound. before reproduction.